Estimating transaction risk using sub-models characterizing cross-interaction among categorical and non-categorical variables

An article comprising a non-transitory machine-readable medium embodying instructions that when performed by one or more machines result in operations comprising:

• receiving input data for a predictive model comprising values for categorical variables and one or more of binary variables and continuous variables, the predictive model characterizing a level of risk for a data transaction;
• associating one or more of the categorical variables with one of a plurality of keys, each key having corresponding coefficients for at least a subset of the binary variables and the continuous variables, the coefficients being dependent on a value for the key;
• generating composite values for composite variables based on values for each of at least a subset of the binary variables and the continuous variables as calculated using the corresponding coefficients for each key;
• and
• initiating scoring of the data transaction by the predictive model using the binary variables, the continuous variables, and the composite variables.

A computer-implemented method comprising:

• receiving input data for a predictive model comprising values for categorical variables and one or more of binary variables and continuous variables, the predictive model characterizing a level of risk for a data transaction;
• associating, for each of at least two cascading sub-models, one or more of the categorical variables with one of a plurality of keys, each key having corresponding coefficients for at least a subset of the binary variables and the continuous variables, the coefficients being dependent on a value for the corresponding key, output for at least one of the cascading sub-models acting as a direct or indirect input to at least one of the other cascading sub-models;
• generating composite values for composite variables based on values for each of at least a subset of the binary variables and the continuous variables as calculated using the corresponding coefficients for each key;
• and
• initiating scoring of the data transaction by the predictive model using the binary variables, the continuous variables, and the composite variables determined for each of the at least two cascading sub-models.